Abstract
Albino rats were trained in a 2-choice probability learning situation for 1,000 trials. A 4 X 4 X 2 factorial design was used incorporating incentive values of 6, 12, 24, and 48% concentrations of a sucrose solution, probabilities of reinforcement of .60, .70, .80, and .90, and a noncorrection vs. a correction training procedure. Results showed terminal response probability to be: (a) higher under noncorrection than under correction, (b) independent of incentive, (c) an increasing function of probability of reinforcement Rate of learning was more rapid with higher incentives with the non-correction procedure but not with correction. All groups exceeded the asymptotic response probability predicted by statistical learning theory. It was concluded that something near 1,000 trials is necessary to accurately assess behavior in probabilistic outcome situations.

This publication has 3 references indexed in Scilit: